Fault diagnosis method, device and equipment for large die forging press and medium

Abstract

The invention discloses a fault diagnosis method and device for a large die forging press, equipment and a medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining multi-modal sensor data of the large die forging press; the data of the multi-mode sensor are preprocessed; inputting the preprocessed multi-modal sensor data into a multi-modal attention convolutional neural network model for feature extraction to obtain multi-modal features; carrying out fusion processing on the multi-modal features; and inputting the fused multi-modal features into a classifier for classification to obtain a fault diagnosis result.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a fault diagnosis method, device, equipment and medium for a large die forging press. Background Art

[0002] The continuous advancement of intelligence and automation in high-end industrial equipment places higher demands on the safe, reliable operation and efficient maintenance of these equipment. Large die forging presses, as essential manufacturing equipment for key metal components, play a vital role in the fields of machinery, automobiles, ships, aerospace, and more. However, due to the constant influence of environmental forces such as temperature, pressure, and impact during the production process, die forging presses frequently malfunction, leading to a gradual decline in equipment efficiency and product quality, resulting in significant economic losses for enterprises. Therefore, fault diagnosis technology has become a key means of ensuring the safe and reliable operation of large die forging presses.

[0003] Currently, fault diagnosis technologies for large die forging presses can be divided into two main areas: those based on mechanistic models or manual experience, and those based on machine learning and deep learning. Mechanism-based methods typically rely on a deep understanding of the equipment's internal mechanisms, constructing mathematical models to predict and diagnose faults. Machine learning and deep learning methods, on the other hand, utilize extensive historical data to train models and implement fault diagnosis by learning the data's characteristics and patterns.

[0004] While existing fault diagnosis technologies have improved the fault diagnosis capabilities of large die forging presses to a certain extent, significant flaws remain. For methods based on mechanism models, since large die forging presses are typically composed of numerous components, the inherent mechanisms are difficult to determine. The relationships between different components are complex and highly coupled, resulting in high costs for traditional fault diagnosis based on mechanism models or manual experience, and their inability to function effectively in actual production processes. As for methods based on machine learning and deep learning, most current examples rely on single-sensor fault prediction and classification, failing to consider the data dependencies between multiple sensors. This results in incomplete and inaccurate diagnostic results. Summary of the Invention

[0005] The purpose of this application is to provide a fault diagnosis method, device, equipment and medium for a large die forging press.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a fault diagnosis method for a large die forging press, comprising:

[0008] Acquire multimodal sensor data from a large die forging press;

[0009] Preprocessing the multimodal sensor data;

[0010] The preprocessed multimodal sensor data is input into the multimodal attention convolutional neural network model for feature extraction to obtain multimodal features;

[0011] Performing fusion processing on the multimodal features;

[0012] The fused multimodal features are input into the classifier for classification to obtain the fault diagnosis results.

[0013] Optionally, the step of acquiring multimodal sensor data of a large die forging press includes:

[0014] Data are collected synchronously from the displacement sensor, impact force sensor, air pressure sensor and hydraulic pressure sensor in the large die forging press to obtain multimodal sensor data.

[0015] Optionally, the step of preprocessing the multimodal sensor data includes:

[0016] The multimodal sensor data is smoothed, filtered, and normalized.

[0017] Optionally, the step of inputting the preprocessed multimodal sensor data into a multimodal attention convolutional neural network model for feature extraction to obtain multimodal features includes:

[0018] Build a multimodal attention convolutional neural network model with multiple input channels and attention mechanisms;

[0019] Inputting the preprocessed multimodal sensor data into each input channel of the modality attention convolutional neural network;

[0020] The weights of the channel data in each of the input channels are learned through the attention mechanism, and feature extraction is performed to obtain multimodal features.

[0021] Optionally, the step of fusing the multimodal features includes:

[0022] Performing dimensionality reduction processing on the multimodal features extracted by the multimodal attention convolutional neural network model to obtain a low-dimensional feature vector;

[0023] The low-dimensional feature vectors are fused.

[0024] Optionally, the step of inputting the fused multimodal features into a classifier for classification to obtain a fault diagnosis result includes:

[0025] Build a TSVM classifier;

[0026] The fused multimodal features are input into the TSVM classifier for training;

[0027] The trained transductive support vector machine is used to classify unknown fault samples and obtain the fault diagnosis results.

[0028] Optionally, the method further includes:

[0029] Before training the multimodal attention convolutional neural network model, labeled samples are used for pre-training.

[0030] In a second aspect, the present application provides a fault diagnosis device for a large die forging press, comprising:

[0031] A data preparation module for acquiring multimodal sensor data from a large die forging press;

[0032] Preprocessing the multimodal sensor data;

[0033] The feature extraction module is used to input the preprocessed multimodal sensor data into the multimodal attention convolutional neural network model for feature extraction to obtain multimodal features;

[0034] Performing fusion processing on the multimodal features;

[0035] The prediction module is used to input the fused multimodal features into the classifier for classification to obtain the fault diagnosis results.

[0036] Optionally, the data preparation module is further configured to:

[0037] Data are collected synchronously from the displacement sensor, impact force sensor, air pressure sensor and hydraulic pressure sensor in the large die forging press to obtain multimodal sensor data.

[0038] Optionally, the data preparation module is further configured to:

[0039] The multimodal sensor data is smoothed, filtered, and normalized.

[0040] Optionally, the feature extraction module is used to:

[0041] Build a multimodal attention convolutional neural network model with multiple input channels and attention mechanisms;

[0042] Inputting the preprocessed multimodal sensor data into each input channel of the modality attention convolutional neural network;

[0043] The weights of the channel data in each of the input channels are learned through the attention mechanism, and feature extraction is performed to obtain multimodal features.

[0044] Optionally, the feature extraction module is further configured to:

[0045] Performing dimensionality reduction processing on the multimodal features extracted by the multimodal attention convolutional neural network model to obtain a low-dimensional feature vector;

[0046] The low-dimensional feature vectors are fused.

[0047] Optionally, the prediction module is further configured to:

[0048] Build a TSVM classifier;

[0049] The fused multimodal features are input into the TSVM classifier for training;

[0050] The trained transductive support vector machine is used to classify unknown fault samples and obtain the fault diagnosis results.

[0051] Optionally, the prediction module is further configured to:

[0052] Before training the multimodal attention convolutional neural network model, labeled samples are used for pre-training.

[0053] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the fault diagnosis method for a large die forging press as described above.

[0054] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the fault diagnosis method for a large die forging press as described above.

[0055] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for diagnosing the faults of large die forging presses.

[0056] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0057] The present application provides a fault diagnosis method, apparatus, equipment and medium for a large die forging press. Through multimodal sensor data fusion and preprocessing, a multimodal attention convolutional neural network is used to automatically extract key features, thereby overcoming the problem of limited information from a single sensor. By performing feature fusion, the data dimension and computational complexity are reduced while retaining the most critical feature information. The fused features are input into a classifier for classification, fully utilizing unlabeled data, and significantly improving the accuracy and generalization ability of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 A schematic flow chart of a fault diagnosis method for a large die forging press provided in one embodiment of the present application;

[0060] Figure 2 This is one of the principle schematic diagrams of a fault diagnosis method for a large die forging press provided in one embodiment of the present application;

[0061] Figure 3 This is a second schematic diagram of the principle of a fault diagnosis method for a large die forging press provided in one embodiment of the present application;

[0062] Figure 4 This is a third schematic diagram of a principle of a fault diagnosis method for a large die forging press provided in one embodiment of the present application;

[0063] Figure 5 A schematic diagram of the functional modules of a fault diagnosis device for a large die forging press provided in one embodiment of the present application;

[0064] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] like Figure 1 As shown, some embodiments of the present application provide a fault diagnosis method for a large die forging press, which includes the following steps 101 to 104. Among them:

[0067] Step 101: Acquire multimodal sensor data of a large die forging press.

[0068] In the embodiments of this application, a large die forging press, as equipment for manufacturing key metal components, has its operating state affected by multiple physical parameters, such as displacement, impact force, air pressure, hydraulic pressure, etc. This data is referred to as multimodal sensor data because it comes from different types of sensors and reflects the operating state of the equipment in different aspects.

[0069] Specifically, based on the needs of fault diagnosis, appropriate sensor types are selected, such as displacement sensors, pressure sensors, and temperature sensors. Data is collected in real time or at scheduled intervals through the sensor interface to ensure timeliness and accuracy. The collected data is stored in a database for subsequent processing and analysis.

[0070] Step 102: pre-process the multimodal sensor data.

[0071] In the embodiments of this application, because data directly acquired from sensors may contain noise, outliers, and inconsistent data formats, preprocessing is required to improve data quality. Preprocessing steps include data cleaning, data smoothing, and data normalization to ensure that the data input to the model is clean, accurate, and in a consistent format.

[0072] Specifically, outliers and missing values ​​are removed or corrected to ensure data integrity. Noise in the data is reduced through filtering algorithms (such as moving average filtering and Kalman filtering). Data of different dimensions are converted to the same dimension to facilitate subsequent processing.

[0073] Step 103: Input the preprocessed multimodal sensor data into a multimodal attention convolutional neural network model to extract features and obtain multimodal features.

[0074] In the present embodiment, a multimodal attention convolutional neural network (MACNN) model is used to extract features from the preprocessed data. The MACNN model combines a convolutional neural network (CNN) with an attention mechanism to automatically learn and extract features relevant to fault diagnosis from the data.

[0075] Specifically, we first designed and built a MACNN model, consisting of an input layer, convolutional layers, attention mechanism layers, and pooling layers. The preprocessed data was then fed into the MACNN model, where multimodal features were extracted through convolution operations and the attention mechanism. The attention mechanism automatically adjusts the weights of different sensor data, focusing on features that are more important for fault diagnosis.

[0076] Reference Figure 2The model consists of two main components: a convolutional neural network (MACNN) with a multimodal attention mechanism, and a PCA-TSVM classifier. The entire model is based on semi-supervised learning. However, since traditional CNN models are typically used for supervised learning, the MACNN component is pre-trained with supervised learning to better extract sample features. This allows the MACNN model to better learn the characteristic representation of labeled samples. This pre-trained MACNN model is then integrated into the overall network to construct the MACNN-PTSVM model.

[0077] from Figure 2 As can be seen from the figure, the backbone of this model can be viewed as a traditional CNN model structure. This structure serves two main purposes: pre-training and feature extraction. Feature extraction primarily consists of convolutional layers, ReLU layers, and max pooling layers. These three layers are stacked twice. The output layer of a conventional CNN model is removed, and PCA is added to perform feature fusion, facilitating fault classification using the TSVM. In this model, a batch normalization layer is used to perform batch normalization on the input before feature extraction in the convolutional layers. This aims to reduce the distribution differences between data collected by different sensors.

[0078] Step 104: performing fusion processing on the multimodal features.

[0079] In the embodiment of the present application, after extracting multimodal features, these features need to be fused to integrate information from different sensor data and improve the accuracy of fault diagnosis. Feature fusion can adopt various methods, such as feature splicing and principal component analysis (PCA).

[0080] Multimodal information fusion is a technique that combines multiple modal information from different sensors to improve recognition accuracy and robustness. In fault diagnosis tasks, multimodal information fusion methods can effectively combine multiple data modalities to improve the model's diagnostic performance.

[0081] For example, Figure 3 As shown in the figure, the 12-channel sensor data is fused by feature fusion, and a 12-channel data feature set is constructed, as shown in Figure 4 shown.

[0082] The system directly concatenates features from different modalities to form a high-dimensional feature vector. Dimensionality reduction is then performed on these concatenated high-dimensional features to remove redundant information and retain the most critical features. PCA projects the data into a new coordinate system through a linear transformation, maximizing the variance (i.e., the most information) in a particular direction within the new coordinate system, thereby preserving the most important features.

[0083] Step 105: Input the fused multimodal features into a classifier for classification to obtain a fault diagnosis result.

[0084] In an embodiment of the present application, the fused multimodal features are input into a classifier for classification to determine whether a large die forging press has a fault and the type of fault. The classifier can use a variety of machine learning algorithms, such as support vector machines (SVM), random forests, etc. Among them, the classifier can be a suitable classifier selected according to the needs of fault diagnosis, such as transduction support vector machines (TSVM), etc. TSVM is a semi-supervised learning algorithm that can be trained using unlabeled data to improve the generalization ability of the model.

[0085] The classifier is then trained using fault data with known labels, enabling it to learn the mapping between fault features and fault types. The fused multimodal features are input into the trained classifier to obtain the fault diagnosis results. The classifier outputs the corresponding fault type or fault probability based on the input features.

[0086] The embodiment of the present application overcomes the problem of limited information from a single sensor by fusing and preprocessing multimodal sensor data and using a multimodal attention convolutional neural network to automatically extract key features. By performing feature fusion, the data dimension and computational complexity are reduced while retaining the most critical feature information. The fused features are input into a classifier for classification, making full use of unlabeled data and significantly improving the accuracy and generalization ability of fault diagnosis.

[0087] Optionally, the step 101 includes: synchronously collecting data from a displacement sensor, an impact force sensor, an air pressure sensor, and a hydraulic pressure sensor in the large die forging press to obtain multimodal sensor data.

[0088] In the embodiment of the present application, the displacement sensor is used to measure the displacement changes of various components or key points of a large die forging press during operation. Displacement data is an important basis for evaluating the operating status of the equipment and predicting potential failures. The impact force sensor is installed on the striking component of the die forging press to monitor the force changes during the striking process. The impact force data can reflect the response of the equipment when it is impacted, and is particularly critical for diagnosing failures caused by excessive impact or abnormal loads. The air pressure sensor monitors the pressure changes of the air pressure system in the die forging press. The stability of the air pressure system is crucial to the normal operation of the equipment, and abnormal air pressure data may indicate a failure of the air pressure system. The hydraulic sensor is used to detect parameters such as pressure and flow of the hydraulic system. As an important component of the die forging press, the operating status of the hydraulic system directly affects the overall performance of the equipment.

[0089] Synchronous data acquisition involves simultaneously acquiring data from the various sensor types mentioned above, ensuring that the collected data remains consistent over time. This is because the operation of a die forging press is a complex dynamic process, with the state changes of various components being interrelated. Through synchronized data acquisition, a complete data set reflecting the real-time operating status of the equipment can be obtained.

[0090] Multimodal sensor data refers to a collection of data collected by multiple different types of sensors. This data contains information on different aspects of the equipment's operation, such as displacement, impact force, air pressure, and hydraulic pressure. Together, these different modal data provide a comprehensive description of the die forging press's operating status.

[0091] By synchronously collecting data from multiple sensors, the embodiments of this application can obtain richer and more comprehensive information about the operating status of equipment. This provides strong support for improving the accuracy and reliability of fault diagnosis. Furthermore, the fusion of multimodal data also provides more dimensional input features for subsequent data analysis and model training, helping to improve the performance of fault diagnosis models.

[0092] Optionally, step 102 includes: performing smoothing, filtering, and normalization on the multimodal sensor data.

[0093] In the embodiments of the present application, the primary purpose of smoothing is to reduce noise and random fluctuations in the data, thereby smoothing the data trend and facilitating subsequent analysis and processing. Common smoothing methods include moving average and Savitzky-Golay filtering. These methods perform weighted averaging or polynomial fitting on the data within a specific window to produce a smoothed data sequence.

[0094] Filtering removes high-frequency noise and interference from data while retaining useful low-frequency information. This is crucial for improving the signal-to-noise ratio (SNR) and highlighting fault characteristics. Common filtering methods include low-pass filtering and band-pass filtering. These methods use a suitable cutoff frequency to filter out signal components above or below that frequency.

[0095] Normalization involves scaling data so that it falls within a small, specific interval (such as [0, 1] or [-1, 1]). This helps eliminate the impact of different dimensions on data analysis and improves the convergence speed and stability of algorithms. Common normalization methods include min-max normalization and Z-score normalization. Min-max normalization linearly transforms data to the interval [0, 1], while Z-score normalization transforms data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0096] By smoothing, filtering, and normalizing multimodal sensor data, the present embodiment can effectively improve data quality, laying a solid foundation for subsequent feature extraction and fault diagnosis. Smoothing and filtering can remove noise and interference signals from the data, highlighting fault characteristics; while normalization eliminates the influence of different dimensions on data analysis, improving the algorithm's convergence speed and stability. These preprocessing steps collectively ensure the accuracy and reliability of subsequent fault diagnosis.

[0097] Optionally, step 103 includes:

[0098] Step 1031, construct a multimodal attention convolutional neural network model including multiple input channels and attention mechanism.

[0099] Step 1032: Input the preprocessed multimodal sensor data into each input channel of the modality attention convolutional neural network.

[0100] Step 1033: learn the weights of the channel data in each of the input channels through the attention mechanism, and perform feature extraction to obtain multimodal features.

[0101] In an embodiment of the present application, a convolutional neural network model capable of processing multimodal data (such as displacement, impact force, air pressure, hydraulic pressure and other multi-sensor data) is built. By introducing an attention mechanism, the model can automatically learn the importance of different modal data to fault diagnosis, thereby improving the accuracy and robustness of fault diagnosis. The model includes multiple input channels, each input channel corresponds to sensor data of a modality. The model also includes a convolution layer, an attention mechanism layer, a pooling layer, etc., for extracting features, learning weights and dimensionality reduction. The attention mechanism is a technology that simulates the human visual attention mechanism, which can automatically pay attention to important information when processing a large amount of information and ignore unimportant information. In a multimodal attention convolutional neural network, the attention mechanism is used to learn the weights of data in different input channels, that is, the contribution of different modal data to fault diagnosis.

[0102] Before inputting data into the model, the raw sensor data must be preprocessed, including smoothing, filtering, and normalization, to improve data quality and model training effectiveness. The preprocessed multimodal sensor data is then fed into the various input channels of the multimodal attention convolutional neural network model. Because data from different modalities may have different physical meanings and dimensions, they must be fed into separate channels.

[0103] During model training, the attention mechanism automatically learns weights for data in different input channels based on the characteristics and label information of the input data. A larger weight indicates a greater contribution of the data in that channel to fault diagnosis; a smaller weight indicates a smaller contribution. Feature extraction is performed on the input data through network structures such as convolutional layers and pooling layers. During feature extraction, the attention mechanism weights the data from different channels based on the learned weights, resulting in more accurate and effective multimodal features.

[0104] This embodiment of the application constructs a multimodal attention convolutional neural network model that includes multiple input channels and an attention mechanism, and inputs preprocessed multimodal sensor data into it to effectively extract multimodal features relevant to fault diagnosis. The introduction of the attention mechanism enables the model to automatically learn the importance of data in different modalities, thereby improving the accuracy and robustness of fault diagnosis.

[0105] Optionally, step 104 includes:

[0106] Step 1041: Perform dimensionality reduction processing on the multimodal features extracted by the multimodal attention convolutional neural network model to obtain a low-dimensional feature vector.

[0107] Step 1042: Fusing the low-dimensional feature vectors.

[0108] In the embodiment of the present application, the multimodal attention convolutional neural network model has completed the feature extraction of the pre-processed multimodal sensor data. These features contain information from different sensors (such as displacement sensors, impact force sensors, air pressure sensors, hydraulic sensors, etc.), which together constitute the multimodal features. The main purpose of dimensionality reduction is to reduce the dimension of the feature vector, thereby simplifying the subsequent calculation process, while removing redundant information and retaining the most critical features. This is of great significance for improving the generalization ability of the model and reducing overfitting. Commonly used dimensionality reduction methods include principal component analysis (PCA), linear discriminant analysis (LDA), t-distributed random neighbor embedding (t-SNE), etc. PCA dimensionality reduction technology converts the original data into a new coordinate system through linear transformation, so that the variance on the first coordinate axis of the new coordinate system is the largest (that is, the projection of the data in this direction is the most dispersed), and so on, thereby achieving data dimensionality reduction. After PCA dimensionality reduction processing, the multimodal features are converted into low-dimensional feature vectors. These low-dimensional feature vectors still retain most of the key information in the original data, but the dimension is significantly reduced.

[0109] Feature fusion is the process of combining features from different sources or modalities into a unified representation, aiming to improve model performance and robustness. Feature fusion leverages the complementary information between data from different modalities, resulting in a more comprehensive and accurate feature representation. There are various methods for feature fusion, including simple concatenation, weighted summation, and kernel methods. A common approach is to concatenate or perform weighted summation on the reduced low-dimensional feature vectors to obtain the final fused features. Feature fusion enables the model to comprehensively utilize information from different sensors, improving the accuracy and comprehensiveness of fault diagnosis. Furthermore, by fusing data from multiple modalities, the model is more tolerant to single sensor failures.

[0110] The present invention uses a multimodal attention convolutional neural network model to extract multimodal features through dimensionality reduction and feature fusion, resulting in a more concise, comprehensive, and accurate feature representation. These features not only retain the key information in the original data but also eliminate redundancy and noise, providing strong support for subsequent classification or regression tasks.

[0111] Optionally, step 105 includes:

[0112] Step 1051: construct a TSVM classifier.

[0113] In the embodiment set of the present application, transduction support vector machine (TSVM) is a semi-supervised learning algorithm, which is particularly suitable for processing situations where there are fewer labeled samples and more unlabeled samples. In the fault diagnosis of large die forging presses, TSVM becomes an effective classification method because it is difficult to obtain a large number of labeled fault samples. The construction process of TSVM usually includes steps such as selecting a kernel function, setting a penalty parameter, and determining an optimization algorithm. The choice of kernel function determines the mapping method of the sample in the feature space. Common kernel functions include linear kernel, polynomial kernel, radial basis kernel (RBF), etc. The penalty parameter is used to balance the complexity and classification error rate of the classifier. The optimization algorithm is used to solve the objective function of TSVM to obtain the optimal classification hyperplane.

[0114] Step 1052: Input the fused multimodal features into the TSVM classifier for training.

[0115] In an embodiment of the present application, during the training phase, the fused multimodal features need to be provided as input data to the TSVM classifier. These features have been subjected to multimodal extraction, dimensionality reduction processing and feature fusion through the previous steps, and can fully reflect the operating status of the large die forging press. The TSVM classifier is trained using input training samples (including a small number of labeled samples and a large number of unlabeled samples). During the training process, TSVM will try to find an optimal classification hyperplane to minimize the classification error rate, while making the best use of the information of unlabeled samples to improve the generalization ability of the classifier.

[0116] Step 1053: Use the trained transductive support vector machine to classify the unknown fault samples to obtain a fault diagnosis result.

[0117] In the embodiment of the present application, after training is completed, the TSVM classifier has the ability to classify unknown fault samples. When a new fault sample appears, it can be input into the trained TSVM classifier, and the classifier will classify it into the corresponding fault category based on the characteristics of the sample and the previously learned classification rules. The output of the TSVM classifier is the fault diagnosis result, that is, the fault category to which the unknown fault sample belongs. This result can be used to guide subsequent repair and maintenance work, helping to ensure the normal operation of the large die forging press.

[0118] By constructing a TSVM classifier and training it with fused multimodal features, this embodiment of the application can produce a classification model that can effectively identify large die forging press faults. This model can fully utilize limited labeled samples and a large amount of unlabeled sample information to improve the accuracy and generalization of fault diagnosis.

[0119] Optionally, the method further includes: before training the multimodal attention convolutional neural network model, using labeled samples for pre-training.

[0120] In an embodiment of the present application, due to the limited number of fault samples of large die forging presses, directly using these limited samples to train a complex convolutional neural network model may lead to problems of overfitting or insufficient model generalization ability. Therefore, before formal training, pre-training the model with labeled samples can help the model initially learn the characteristics and patterns of the data, thereby improving the performance of the model. Pre-training can also help the model converge to the optimal solution faster during the formal training phase. Through pre-training, the model already has certain initial weights and biases, which are closer to the optimal solution, so during formal training, the model can adjust these parameters more quickly to adapt to new data.

[0121] Before pre-training, a certain number of labeled samples must be prepared. These samples should be representative and reflect the data characteristics of a large die forging press under different operating conditions. The quantity and quality of the labeled samples will directly affect the effectiveness of pre-training. Based on the complexity of the model and the number of labeled samples, appropriate pre-training parameters, such as the learning rate, batch size, and number of training rounds, should be set. These parameters will affect the speed and effectiveness of pre-training. The labeled samples are input into the multimodal attention convolutional neural network model and trained according to the preset parameters. During training, the model continuously adjusts its weights and biases to minimize the loss function value. When the loss function value converges to a certain level, pre-training is considered complete.

[0122] The embodiments of the present application can effectively improve the performance and generalization ability of the multimodal attention convolutional neural network model by pre-training with labeled samples before training the model. Pre-training can help the model initially learn the characteristics and patterns of the data, providing a better starting point for formal training. At the same time, pre-training can also accelerate the convergence of the model and improve training efficiency.

[0123] Based on the same inventive concept, embodiments of the present application also provide a large die forging press fault diagnosis device for implementing the aforementioned large die forging press fault diagnosis method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the large die forging press fault diagnosis device provided below can be found in the aforementioned limitations of the large die forging press fault diagnosis method, and will not be further elaborated here.

[0124] In an exemplary embodiment, Figure 5 As shown, a fault diagnosis device 20 for a large die forging press is provided, comprising:

[0125] Data preparation module 201, for acquiring multimodal sensor data of a large die forging press;

[0126] Preprocessing the multimodal sensor data;

[0127] A feature extraction module 202 is used to input the pre-processed multimodal sensor data into a multimodal attention convolutional neural network model to extract features to obtain multimodal features;

[0128] Performing fusion processing on the multimodal features;

[0129] The prediction module 203 is used to input the fused multimodal features into a classifier for classification to obtain a fault diagnosis result.

[0130] Optionally, the data preparation module 201 is further configured to:

[0131] Data are collected synchronously from the displacement sensor, impact force sensor, air pressure sensor and hydraulic pressure sensor in the large die forging press to obtain multimodal sensor data.

[0132] Optionally, the data preparation module 201 is further configured to:

[0133] The multimodal sensor data is smoothed, filtered, and normalized.

[0134] Optionally, the feature extraction module 202 is configured to:

[0135] Build a multimodal attention convolutional neural network model with multiple input channels and attention mechanisms;

[0136] Inputting the preprocessed multimodal sensor data into each input channel of the modality attention convolutional neural network;

[0137] The weights of the channel data in each of the input channels are learned through the attention mechanism, and feature extraction is performed to obtain multimodal features.

[0138] Optionally, the feature extraction module 202 is further configured to:

[0139] Performing dimensionality reduction processing on the multimodal features extracted by the multimodal attention convolutional neural network model to obtain a low-dimensional feature vector;

[0140] The low-dimensional feature vectors are fused.

[0141] Optionally, the prediction module 203 is further configured to:

[0142] Build a TSVM classifier;

[0143] The fused multimodal features are input into the TSVM classifier for training;

[0144] The trained transductive support vector machine is used to classify unknown fault samples and obtain the fault diagnosis results.

[0145] Optionally, the prediction module 203 is further configured to:

[0146] Before training the multimodal attention convolutional neural network model, labeled samples are used for pre-training.

[0147] The embodiment of the present application overcomes the problem of limited information from a single sensor by fusing and preprocessing multimodal sensor data and using a multimodal attention convolutional neural network to automatically extract key features. By performing feature fusion, the data dimension and computational complexity are reduced while retaining the most critical feature information. The fused features are input into a classifier for classification, making full use of unlabeled data and significantly improving the accuracy and generalization ability of fault diagnosis.

[0148] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store information interaction data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a fault diagnosis method for a large die forging press is implemented.

[0149] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0150] In an exemplary embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0151] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0152] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0154] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0155] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0156] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0157] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A fault diagnosis method for a large die forging press, characterized in that: The fault diagnosis method of the large die forging press comprises: Acquire multimodal sensor data from a large die forging press; Preprocessing the multimodal sensor data; The preprocessed multimodal sensor data is input into the multimodal attention convolutional neural network model for feature extraction to obtain multimodal features; Performing fusion processing on the multimodal features; The fused multimodal features are input into the classifier for classification to obtain the fault diagnosis results.

2. The fault diagnosis method for a large die forging press according to claim 1, characterized in that: The step of acquiring multimodal sensor data of a large die forging press comprises: Data is collected synchronously from the displacement sensor, the impact force sensor, the air pressure sensor and the hydraulic pressure sensor in the large die forging press to obtain multi-modal sensor data.

3. The fault diagnosis method for a large die forging press according to claim 2, characterized in that: The step of preprocessing the multimodal sensor data comprises: The multimodal sensor data is smoothed, filtered and normalized.

4. The fault diagnosis method for a large die forging press according to claim 2, characterized in that: The step of inputting the preprocessed multimodal sensor data into a multimodal attention convolutional neural network model for feature extraction to obtain multimodal features includes: Build a multimodal attention convolutional neural network model with multiple input channels and attention mechanisms; Inputting the preprocessed multimodal sensor data into each input channel of the modality attention convolutional neural network; The weights of the channel data in each of the input channels are learned through the attention mechanism, and feature extraction is performed to obtain multimodal features.

5. The fault diagnosis method for a large die forging press according to claim 1, characterized in that: The step of fusing the multimodal features comprises: Performing dimensionality reduction processing on the multimodal features extracted by the multimodal attention convolutional neural network model to obtain a low-dimensional feature vector; The low-dimensional feature vectors are fused.

6. The fault diagnosis method for a large die forging press according to claim 1, characterized in that: The step of inputting the fused multimodal features into a classifier for classification to obtain a fault diagnosis result comprises: Build TSVM classifier; The fused multimodal features are input into the TSVM classifier for training; The trained transductive support vector machine is used to classify unknown fault samples and obtain the fault diagnosis results.

7. The fault diagnosis method for a large die forging press according to claim 6, characterized in that: The method further comprises: Before training the multimodal attention convolutional neural network model, pre-training is performed using labeled samples.

8. A fault diagnosis device for a large die forging press, characterized in that: The fault diagnosis device of the large die forging press comprises: Data preparation module for acquiring multimodal sensor data from large die forging presses; Preprocessing the multimodal sensor data; A feature extraction module is used to input the preprocessed multimodal sensor data into the multimodal attention convolutional neural network model for feature extraction to obtain multimodal features; Performing fusion processing on the multimodal features; The prediction module is used to input the fused multimodal features into the classifier for classification to obtain the fault diagnosis results.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the fault diagnosis method for a large die forging press as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fault diagnosis method for a large die forging press according to any one of claims 1 to 7 are implemented.

Images